‘Big data’ from online recruitment platforms show discrimination against ethnic minorities and women – and sometimes men

We are optimistic that at least part of the discrimination that we document in this study can be overcome by re-designing recruitment platforms.
- Dr Dominik Hangartner
Person holding pencil near laptop Photo by Scott Graham on Unsplash

Recruiters using an online recruitment platform were up to 19 per cent less likely to follow up with job seekers from immigrant and ethnic minority backgrounds than with equally qualified job seekers from the majority population, according to research published in the journal Nature.

The paper ‘Monitoring hiring discrimination through online recruitment platforms’ (1) shows that even when people from immigrant and ethnic minority backgrounds had high levels of work experience, this was not enough to offset this hiring disadvantage.

It also reveals that women were seven per cent less likely to be contacted by recruiters when applying for roles in professions dominated by men. However, the opposite was true for men – who were less likely to be contacted for roles in female dominated industries.

The researchers developed a new method to monitor hiring discrimination by utilising supervised machine learning algorithms(2), which were used to analyse the search behaviour of recruiters using the online recruitment platform of the Swiss public employment service(3). This recruitment platform is similar to JobCentre Plus and also other commercial recruitments websites in the UK.

For the research, data was collected on 452,729 searches by 43,352 recruiters, as well as 17.4 million profiles that appeared in the search lists and 3.4 million profile views. The researchers also looked at the time that recruiters spent looking at each profile and their decisions about whether to contact a jobseeker or not.

They linked the profiles of jobseekers to the unemployment register and showed that every click on the ‘contact candidate button’ by a recruiter increased the probability that an individual got a job in the next three months by two per cent.

Dr Dominik Hangartner, co-author of the paper and Associate Professor in Political Science at the London School of Economics and Political Science (LSE), said: “Our results demonstrate that recruiters treat otherwise identical jobseekers who appear in the same search list differently, depending on their immigrant or minority ethnic background. Unsurprisingly, this has a real impact on who gets employed.”

The researchers found that there were only very small differences in the time spent by recruiters on the profiles of individuals from immigrant and minority ethnic groups relative to those from the majority population, showing that it is unlikely that recruiters use ethnicity as a shortcut to screen out applicants.

However, at certain times of the day – just before lunch (11.00 -11.59am) or towards the end of the work day (5.00-5.59pm) – recruiters spent less time looking at all CVs.  During these hours, when recruiters reviewed faster, immigrant and minority ethnic groups faced up to 20 per cent higher levels of discrimination.

Dr Hangartner said: “These results suggest that unconscious biases, such as stereotypes about minorities, have a larger impact when recruiters are more tired and fall back on ‘intuitive decision-making’.

“We are optimistic that at least part of the discrimination that we document in this study can be overcome by re-designing recruitment platforms. For example, more relevant information such as a candidate’s work experience and education could be placed at the top, and details which might indicate ethnicity or gender, such as name or nationality, could appear much lower down the CV.”

Behind the article

(1) ‘Monitoring hiring discrimination through online recruitment platforms’ by Dominik Hangartner (LSE & ETH Zurich), Daniel Kopp (ETH Zurich)  Michael Siegenthaler (ETH Zurich).

(2) This methodology provides a widely applicable, non-intrusive, and cost-efficient tool that researchers and policy-makers can employ to continuously monitor hiring discrimination, to illuminate some of the drivers of discrimination, and to inform approaches to counter it.  Recruitment discrimination is often studied by correspondence tests, where researchers send fictitious resumes that are identical except for the randomised minority trait to be tested (e.g. Black vs. White-sounding names). While this approach is valuable, it can only be usually used to study a few characteristics of applicants in select occupations at a particular point in time. There are also ethical concerns regarding the submission of fictitious applications, which involves deception and increases the workload of human resources personnel and could interfere with the outcome of real applications. In contrast, the methodology introduced in this study examines hiring discrimination by tracking recruiters’ search behaviour of real CVs on an employment website and using machine learning to control for all relevant jobseeker characteristics that are visible to recruiters.

(3) For the purposes of this research, immigrant and ethnic minorities were defined as people who didn’t have Swiss citizenship, who spoke at least one non-Swiss language and whose name is classified as being of non-Swiss origin.